"Felger Carbon" wrote
"Harry Andreas" wrote

Methinks there's some confusion there between processors, avionics
architecture, and software.
While it's true that Intel tried to shut down i960 production
causing a
chinese fire drill, there are enough assets to get by until a new
processor is
ready.

Full disclosu I'm a retired electrical engineer. I specialized in
high-end embedded microprocessors, which the "i960" in the F-22 is. I
know nothing about designing aircraft. I do know a little about the
Intel processor at the heart of the F-22:

The i960MX was designed by Intel specifically and solely for the F-22.
It was an extremely large chip (die) for its day, being octagonal in
shape so that the production 'mask' could be projected by the largest
then-available optics. The reason for the large size was the triple
redundancy built into the chip, which is what separates the i960MX
from all the other COTS i960s. The i960MX was _not_ a particularly
high-performance part, even when new.

It's likely that the reason Intel issued an end-of-life advisory on
the i960MX was that Intel was closing down the last 'fab' that was
capable of running that long-ago technology. A chip with a million
transistors on the die was pretty big back then, while 200 million
transistors per die is routine today.

Intel built the i960MX at the tail end of the period when electronics
companies would manufacture special "mil spec" parts for the military.
I suppose it was intended to be a public-relations gesture, as the
part was most certainly never going to make a profit, or anything
remotely close to that.

I don't think Intel realized that the F-22 project was going to drag
out so long that no significant production would occur before the
manufacturing technique and facilities for the i960MX would become so
obsolete (not merely obsolescent).

These problems exist WRT replacing the i960MX:

1. It unlikely that Intel would, these days, agree to build a special
triple-redundant microprocessor as a replacement. Charity for the
military is now a vanished concept, even (especially?) as a
public-relations effort.

2. You can't just replace a dozen-plus-year-old micro with a new one
and make no other changes. A complete new computer subsystem would
have to be designed. New parts are simply too fast to have any chance
of working in the old system.

3. Airplane controllers are necessarily real-time systems, and that
means a vastly-faster microprocessor, while a good thing at an
abstract level, requires a total re-write of the software (aside from
the new features to be designed in). If the system were _not_
real-time, this could be avoided. Alas.

I now return you to your regularly scheduled aircraft experts. ;-)

Thanks for the insight into the i960MX.

As it happens, GDIS in Mineapolis has developed a quad-redundant voting
PPC603E module for spacecraft applications. Voting schemes aren't new, years
ago Harris build a triple R3000 hybrid. You're right that you just can't
slip a new CPU into an existing system. You_can_build a new processor CCA
compliant with the existing ICD. F-22 uses a multicomputer architecture
based on FDDI which is of course stone slow by current standards. The FDDI
plant should be adequate until the next block change. Newer systems are
using FibreChannel signaling at 1GBaud while 10Gb Ethernet is the future.
That said,

You're quite right about the real-time code for flight control. The RADAR
data and signal processing and ICNIA functions should be fairly unperturbed
and quite portable.